Fault prediction method and device and baseboard management controller

Abstract

The invention provides a fault prediction method and device and a baseboard management controller, and is applied to the technical field of fault prediction, and the method comprises the steps: obtaining the real-time operation data of a server; the real-time operation data comprises operation parameters of the hardware equipment, log data generated in the operation process of the hardware equipment, hardware data of the hardware equipment and software data of the software system. And determining fault prediction characteristics of the server according to the operation parameters, the log data, the hardware data and the software data. The fault prediction features comprise hard disk fault prediction features and memory fault prediction features. And inputting the hard disk fault prediction features and the memory fault prediction features into the trained fault prediction model, and determining a predicted fault type of the server. And on the basis of the predicted fault type, generating a fault diagnosis suggestion, thereby solving the problem that the fault diagnosis efficiency of the server is relatively low due to the fact that a technician needs to perform offline determination when the fault in the server is predicted at present.

Description

Technical Field

[0001] The present application relates to the technical field of fault prediction, and in particular to a fault prediction method, device and baseboard management controller. Background Art

[0002] In modern data centers and cloud computing environments, baseboard management controllers (BMCs) are the core components of servers and play an indispensable role. BMCs are mainly responsible for monitoring and managing servers to ensure that they can operate in the best condition.

[0003] However, the functions of traditional BMCs are mainly focused on data collection and real-time monitoring. The fault diagnosis function that BMCs can achieve is very limited, and it can only determine whether the operating data of the server is abnormal. If more locations and more types of faults in the server are to be predicted, technicians are required to determine them offline, which causes the server fault diagnosis process to lag behind for a long time, and may eventually affect the normal operation of the server due to low fault diagnosis efficiency. Summary of the invention

[0004] The purpose of the present application is to provide a fault prediction method, device and baseboard management controller to solve the problem that when predicting faults in a server, technical personnel are required to make offline determinations, resulting in low efficiency in server fault diagnosis.

[0005] In a first aspect, an embodiment of the present application provides a fault prediction method, which is applied to a baseboard management controller, and the baseboard management controller is used to monitor and manage a server, wherein the server includes a hardware device, and a software system runs on the server, and the method includes: obtaining real-time operation data of the server. The real-time operation data includes the operation parameters of the hardware device, the log data generated by the hardware device during operation, the hardware data of the hardware device, and the software data of the software system. According to the operation parameters, log data, hardware data and software data, the fault prediction characteristics of the server are determined. The fault prediction characteristics include hard disk fault prediction characteristics and memory fault prediction characteristics. The hard disk fault prediction characteristics and memory fault prediction characteristics are input into the trained fault prediction model to determine the predicted fault type of the server. Based on the predicted fault type, a fault diagnosis suggestion is generated.

[0006] In a possible implementation, the training process of the fault prediction model includes: obtaining historical operation data of the server. The historical operation data carries a fault type label. The historical operation data includes historical operation parameters of the hardware device, historical log data of the hardware device, historical hardware data of the hardware device, and historical software data of the software system. According to the historical operation parameters, historical log data, historical hardware data, and historical software data, the fault characteristics of the server are determined. The fault characteristics include hard disk fault characteristics and memory fault characteristics. According to the hard disk fault characteristics and memory fault characteristics, the fault prediction model is trained.

[0007] In a possible implementation, hard disk failure characteristics include SMART data characteristics, read / write speed characteristics, and run time characteristics. Memory failure characteristics include memory statistical characteristics, time series characteristics, and frequency domain characteristics. According to historical operating parameters, historical log data, historical hardware data, and historical software data, determining the failure characteristics of the server includes: screening out SMART parameters, read / write speed parameters, run time parameters, memory statistical parameters, time series parameters, and frequency domain parameters from historical operating parameters, historical log data, historical hardware data, and historical software data. According to SMART parameters and SMART change trends of SMART parameters in a preset time period, determining SMART data characteristics. According to read / write speed parameters and read / write speed change trends of read / write speed parameters in a preset time period, determining read / write speed characteristics. According to run time parameters, determining run time characteristics. Performing statistical processing on memory statistical parameters within a preset time period to obtain memory statistical characteristics. Performing feature extraction on time series parameters based on a sliding window to obtain time series characteristics. Performing Fourier transform on frequency domain parameters to obtain frequency domain characteristics.

[0008] A possible implementation method is to train a fault prediction model based on hard disk fault characteristics and memory fault characteristics, including: using the historical operation data of the server to build a sample set. Configure the number of decision trees of the fault prediction model, the depth of each decision tree, the number of features used by the nodes on each decision tree, the iteration termination condition of the fault prediction model, the number of samples for each node, and the information gain of each node. Extract training samples from the sample set with replacement to form a training set. Use the training set to start training from the root node of each decision tree until all nodes of the decision tree are trained to obtain a fault prediction model. The root node is any node on the decision tree.

[0009] In a possible implementation, the fault prediction method provided in the embodiment of the present application further includes: extracting test samples from the sample set to form a test set. Inputting the test set into the trained fault prediction model to obtain a fault test result. Using the Gini parameter to evaluate the fault test result to obtain the Gini index. Based on the Gini index, adjusting the parameters of the fault prediction model.

[0010] In a possible implementation, the fault prediction method provided in the embodiment of the present application further includes: using interpolation to fill missing values ​​in the real-time operation data. And / or, eliminating or correcting abnormal values ​​in the real-time operation data. And / or, removing duplicate values ​​in the real-time operation data.

[0011] In a possible implementation, the fault prediction method provided in the embodiment of the present application further includes: determining the fault probability and fault level corresponding to the predicted fault type of the server according to the hard disk fault prediction characteristics and the memory fault prediction characteristics. When the fault probability is greater than or equal to the probability threshold, an alarm signal with a fault level is generated. The alarm signal is used to instruct the user to allocate maintenance resources to the server according to the fault level.

[0012] In a second aspect, an embodiment of the present application provides a fault prediction device, which is applied to a baseboard management controller, and the baseboard management controller is used to monitor and manage a server, and the server includes a hardware device, and a software system runs on the server. The device includes: an acquisition module, a determination module, and a generation module.

[0013] The acquisition module is used to acquire the real-time operation data of the server. The real-time operation data includes the operation parameters of the hardware device, the log data generated by the hardware device during operation, the hardware data of the hardware device and the software data of the software system.

[0014] The determination module is used to determine the fault prediction features of the server based on the operating parameters, log data, hardware data and software data. The fault prediction features include hard disk fault prediction features and memory fault prediction features. The hard disk fault prediction features and memory fault prediction features are input into the trained fault prediction model to determine the predicted fault type of the server.

[0015] The generation module is used to generate fault diagnosis suggestions based on the predicted fault type.

[0016] In a third aspect, an embodiment of the present application provides a baseboard management controller, which has the function of implementing the fault prediction method of the first aspect or any possible implementation method. The function can be implemented by hardware, or by hardware executing corresponding software. The hardware or software includes one or more modules corresponding to the above functions.

[0017] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, which stores instructions. When the computer-readable storage medium is run on a computer, the computer can execute the fault prediction method of the first aspect or any possible implementation method.

[0018] In a fifth aspect, an embodiment of the present application provides a computer program product comprising instructions, which, when executed on a computer, enables the computer to execute the fault prediction method of the first aspect or any possible implementation method.

[0019] Among them, the technical effects brought about by any design method in the second to fifth aspects can refer to the technical effects brought about by different possible implementation methods in the first aspect, and will not be repeated here. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] In order to more clearly illustrate the specific implementation methods of the present application or the technical solutions in the prior art, the drawings required for use in the specific implementation methods or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some implementation methods of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0021] Figure 1 A structural diagram of a baseboard management controller provided in an embodiment of the present application;

[0022] Figure 2 A structural schematic diagram of a fault prediction system provided in an embodiment of the present application;

[0023] Figure 3 A flowchart of a fault prediction method provided in an embodiment of the present application;

[0024] Figure 4 A structural schematic diagram of a fault prediction device provided in an embodiment of the present application;

[0025] Figure 5 Another structural schematic diagram of a fault prediction system provided in an embodiment of the present application. DETAILED DESCRIPTION

[0026] In order to make the purpose, technical solution and advantages of the embodiments of the present application clearer, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. The components of the embodiments of the present application described and shown in the drawings here can be arranged and designed in various different configurations.

[0027] Therefore, the following detailed description of the embodiments of the present application provided in the accompanying drawings is not intended to limit the scope of the present application for which protection is sought, but merely represents selected embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in the field without creative work are within the scope of protection of the present application.

[0028] At present, the functions of traditional BMC are mainly concentrated on data collection and real-time monitoring. The fault diagnosis function that BMC can achieve is very limited, and it can only determine whether the operating data of the server is abnormal. If more locations and more types of faults in the server are to be predicted, technicians are required to determine them offline, which causes the server fault diagnosis process to lag behind for a long time, and may eventually affect the normal operation of the server due to low fault diagnosis efficiency.

[0029] Based on this, an embodiment of the present application provides a fault prediction method, which is applied to a baseboard management controller. The baseboard management controller is used to monitor and manage a server. The server includes a hardware device. A software system runs on the server. The method includes: obtaining real-time operation data of the server. The real-time operation data includes the operation parameters of the hardware device, the log data generated by the hardware device during operation, the hardware data of the hardware device, and the software data of the software system. According to the operation parameters, log data, hardware data and software data, the fault prediction characteristics of the server are determined. The fault prediction characteristics include hard disk fault prediction characteristics and memory fault prediction characteristics. The hard disk fault prediction characteristics and memory fault prediction characteristics are input into the trained fault prediction model to determine the predicted fault type of the server. Based on the predicted fault type, a fault diagnosis suggestion is generated.

[0030] The fault prediction method provided in the embodiment of the present application can predict the predicted fault type of the server based on the real-time operation data of the server, and generate fault diagnosis suggestions in advance for the predicted fault type. It can detect potential memory failures and hard disk failures in the server in advance, allowing maintenance personnel to take timely measures to reduce the time and cost required for maintenance.

[0031] On the one hand, the embodiment of the present application provides a baseboard management controller, such as Figure 1 As shown, the baseboard management controller 100 may include a processor 101 , a memory 102 and at least one sensor 103 .

[0032] The processor 101 may include one or more processing units, for example, the processor 101 may include an application processor (AP), a modem processor, a graphics processor (GPU), an image signal processor (ISP), a controller, a memory, a video codec, a digital signal processor (DSP), a baseband processor, and / or a neural network processing unit (NPU). Different processing units may be independent devices or integrated into one or more processors. The processor 101 may also use a low-power, high-performance embedded processor, such as an ARM architecture or a RISC-V architecture.

[0033] Exemplarily, the processor 101 may be specifically responsible for running BMC firmware and processing various management tasks.

[0034] The memory 102 is used to store instructions and data. In some embodiments, the memory 102 may be a cache memory. The memory 102 may store instructions or data that the processor 101 has just used or circulated. If the processor needs to use the instruction or data again, it may be directly called from the memory. This avoids repeated access, reduces the waiting time of the processor 101, and improves the efficiency of the baseboard management controller. The memory 102 may also be a flash memory, a read-only memory (ROM), a random access memory (RAM), etc.

[0035] Sensor 103 can be used to collect hardware status information of the server, such as temperature, voltage, circuit, fan speed, etc. Sensor 103 can be a temperature sensor, a voltage sensor, a current sensor, a fan speed sensor, etc. The application does not limit the type and quantity of sensor 103.

[0036] It should be noted that the above Figure 1 The illustrated baseboard management controller 100 is only used as an example to illustrate the application scenario of the solution of the present application, and is not intended to limit the application scenario of the solution of the present application.

[0037] On the one hand, an embodiment of the present application provides a fault prediction system, and the fault prediction system 200 may include: a baseboard management controller 100 and a server 201 .

[0038] The baseboard management controller 100 and the server 201 may be connected via a variety of interfaces and communication protocols. The baseboard management controller 100 is used to monitor, manage and control the server 201 .

[0039] Exemplarily, the connection between the baseboard management controller 100 and the server 201 may include: 2 C (Inter-Integrated Circuit) bus, SPI (Serial Peripheral Interface) bus, UART (Universal Asynchronous Receiver / Transmitter), IPMI (Intelligent Platform Management Interface), KVM over IP (Keyboard, Video, Mouse over IP), LAN (Local Area Network) interface, PCIe (Peripheral Component Interconnect Express), SMBus (System Management Bus) and BIOS / UEFI interface, etc., which are not limited in this application.

[0040] It should be noted that the above Figure 2 The illustrated fault prediction system 200 is only an example of the application scenario of the present application solution, and is not a limitation on the application scenario of the present application solution. The present application does not limit the number of baseboard management controllers 100 and the number of servers 201 in the fault prediction system 200. The number of baseboard management controllers 100 and the number of servers 201 can be increased or decreased according to the specific fault prediction function requirements.

[0041] On the one hand, the embodiment of the present application provides a fault prediction method, which can be deployed as follows Figure 1 The baseboard management controller 100 shown in FIG. Figure 2 The fault prediction system 200 shown is implemented as follows. Figure 3 As shown, the method includes:

[0042] S301, obtaining real-time operation data of the server.

[0043] Specifically, the baseboard management controller obtains the real-time operation data of the server according to a preset collection frequency.

[0044] The preset collection frequency can be set according to the type of real-time operation data actually collected. For example, for parameters that may change rapidly, such as temperature and current, the collection frequency can be set to once per minute. For indicators that change slowly, such as voltage or hard disk life parameters, the collection frequency can be appropriately reduced, for example, set to once every 10 minutes. This process sets a specific preset collection frequency according to the type of real-time operation data to capture the key time point when the fault occurs, and can also avoid data redundancy and resource consumption caused by too high a frequency.

[0045] The real-time operation data includes the operation parameters of the hardware device, the log data generated by the hardware device during operation, the hardware data of the hardware device and the software data of the software system.

[0046] For example, the key operating parameters of hardware devices are monitored in real time through various sensors built into the BMC. For memory and hard disk, various performance indicators including temperature, voltage, current, etc. are collected. These data can directly reflect the health status of hardware devices and provide basic support for fault prediction and diagnosis.

[0047] By acquiring server log data, we can extract error information related to memory and hard disk (such as ECC check errors), abnormal behavior (such as overheating or performance abnormalities), and potential risk events. This type of data can help discover potential signs of failure and provide more accurate prediction support.

[0048] In addition to memory and hard disk, other hardware and system data are also crucial for comprehensive fault prediction. Expand the collection of hardware status data such as CPU temperature and utilization, network device traffic and latency, power module voltage and power consumption. In addition, at the software level, collect information such as system load, application running status, resource usage, etc.

[0049] Furthermore, after obtaining the real-time operation data of the server, the real-time data is preprocessed.

[0050] Specifically, using interpolation to fill missing values ​​in real-time operation data can ensure the continuity and integrity of real-time operation data. Eliminating or correcting outliers in real-time operation data can avoid the negative impact of outliers in real-time operation data on fault prediction, such as using Z-score or IQR to identify and process outliers. Removing duplicate values ​​in real-time operation data can avoid repeated recording of real-time operation data that causes deviations in the model during training, and ensure the uniqueness of the recorded real-time operation data.

[0051] S302, determining the fault prediction characteristics of the server according to the operating parameters, log data, hardware data and software data.

[0052] Among them, fault prediction features include hard disk fault prediction features and memory fault prediction features. Hard disk fault features include SMART data features, read / write speed features, and run time features. Memory fault features include memory statistical features, time series features, and frequency domain features.

[0053] A possible implementation method is to filter out SMART parameters, read / write speed parameters, run time parameters, memory statistics parameters, time series parameters and frequency domain parameters from historical operating parameters, historical log data, historical hardware data and historical software data.

[0054] The SMART data features are determined based on the SMART parameters and the SMART change trends of the SMART parameters in a preset time period.

[0055] For example, by using the SMART (Self-Monitoring, Analysis and Reporting Technology) indicators of the hard disk, SMART parameters such as hard disk temperature, reallocated sector count, and failure rate can be extracted, which can directly reflect the health of the hard disk. And determine the change pattern of SMART parameters over time within a preset time period, identify the deterioration trend of the hard disk health, and discover potential hardware problems in advance.

[0056] The reading and writing speed characteristics are determined according to the reading and writing speed parameters and the reading and writing speed change trends of the reading and writing speed parameters in a preset time period.

[0057] For example, extracting the read / write speed parameters of the hard disk read / write speed, such as the mean, standard deviation, maximum and minimum values, can describe the performance of the hard disk in different time periods. At the same time, determining the read / write speed change trend of the read / write speed parameters over time within a preset time period can identify possible signs of performance degradation.

[0058] Based on the runtime parameters, runtime characteristics are determined.

[0059] For example, the total operating time of the hard disk can be recorded, which can be used to measure the degree of hard disk aging. The relationship between operating time and failure can be recorded. By combining historical failure data, the impact of operating time on the probability of hard disk failure can be quantified to evaluate the reliability of the hard disk.

[0060] Perform statistical processing on memory statistical parameters within a preset time period to obtain memory statistical features.

[0061] For example, memory statistical parameters related to memory are extracted within a preset time period, such as memory usage, read and write speed, and ECC (error correction code) error rate. Then, the memory statistical parameters within the preset time period are statistically processed, such as calculating the mean, standard deviation, maximum and minimum values, etc. Among them, the mean can reflect the overall level of memory performance. The standard deviation can measure the amplitude of memory performance fluctuations and help identify abnormal fluctuations. The maximum and minimum values ​​can be used to capture extreme usage and identify whether there are performance bottlenecks or anomalies. Memory statistical features can effectively reflect the basic working status and overall health level of memory.

[0062] The time series parameters are extracted based on the sliding window to obtain the time series features.

[0063] For example, based on the sliding window method, feature extraction is performed on time series parameters to calculate dynamic statistical values ​​of performance indicators such as memory usage and read / write speed within a specific time period, such as mean and variance. The time interval in the sliding window can be five minutes or ten minutes. By setting a fixed time interval, changes in memory performance indicators can be dynamically analyzed. This process can identify progressive problems or periodic failure modes by capturing the changes in memory usage over time.

[0064] Perform Fourier transform on the frequency domain parameters to obtain frequency domain features.

[0065] For example, Fourier transform is performed on time series parameters such as memory read and write speed, and the time series parameters are converted from the time domain to the frequency domain to extract frequency-related features. This frequency domain feature can reveal the implicit periodicity in real-time operation data and help identify abnormal signals within a specific frequency range.

[0066] S303: Input the hard disk fault prediction features and the memory fault prediction features into the trained fault prediction model to determine the predicted fault type of the server.

[0067] In a possible implementation, the training process of the fault prediction model can be to obtain the historical operation data of the server. The historical operation data carries a fault type label. The historical operation data includes the historical operation parameters of the hardware device, the historical log data of the hardware device, the historical hardware data of the hardware device, and the historical software data of the software system. According to the historical operation parameters, the historical log data, the historical hardware data, and the historical software data, the fault characteristics of the server are determined. The fault characteristics include the hard disk fault characteristics and the memory fault characteristics. According to the hard disk fault characteristics and the memory fault characteristics, the fault prediction model is trained.

[0068] Specifically, the sample set is constructed using the historical operation data of the server. The number of decision trees of the fault prediction model, the depth of each decision tree, the number of features used by the nodes on each decision tree, the iteration termination condition of the fault prediction model, the number of samples of each node, and the information gain of each node are configured. Training samples are extracted from the sample set with replacement to form a training set. The training is started from the root node of each decision tree using the training set until all nodes of the decision tree are trained to obtain a fault prediction model. The root node is any node on the decision tree.

[0069] Furthermore, test samples are extracted from the sample set to form a test set. The test set is input into the trained fault prediction model to obtain the fault test result. The fault test result is evaluated using the Gini parameter to obtain the Gini index. Based on the Gini index, the parameters of the fault prediction model are adjusted.

[0070] Exemplarily, the fault detection model can select a random forest model. After obtaining the historical operation data of the server, fault features related to fault prediction are extracted from the historical operation data. By calculating the contribution of each fault feature to the split node in the fault prediction model, its importance to the prediction result of the fault prediction model is quantified.

[0071] The Gini index reduction of the fault feature at the split nodes of all trees is counted. The higher the score, the more important the feature. The information gain brought by each feature in the classification task is calculated, and the features with higher information gain are selected as the core input to further optimize the data dimension.

[0072] Furthermore, the data set is divided into training set, validation set and test set. The training set is used for initial training of the model, accounting for 70%-80% of the total data. The validation set is used to adjust the model parameters and structure, accounting for 10%-15% of the total data. The test set is used for final performance evaluation, accounting for 10%-15% of the total data.

[0073] Furthermore, the data set is divided into K equal parts, one of which is selected as the validation set each time, and the remaining K-1 parts are selected as the training set, and the cycle is repeated K times. Finally, the average value of each validation result is taken as the evaluation standard of model performance. Through cross-validation, the hyperparameters of random forest are adjusted, such as the number of decision trees (n_estimators), the maximum depth of a single tree (max_depth), and the minimum number of split samples (min_samples_split), etc., to improve the prediction effect of the model.

[0074] Then use the accuracy, recall and F1 value to evaluate the performance of the fault prediction model on the validation set and test set. Accuracy is used to measure the accuracy of the overall prediction of the model and is suitable for the case where the data category distribution is balanced. Recall is used to evaluate the model's ability to identify positive samples and is suitable for tasks that are sensitive to missing important positive samples. F1 value is used to comprehensively consider the harmonic mean of accuracy and recall, and is suitable for the case where the category distribution is uneven. According to the evaluation results, the model parameters, such as the number of trees, feature selection range, etc., are continuously adjusted to optimize the model performance.

[0075] One possible implementation method is to input the hard disk fault prediction features and memory fault prediction features into the trained fault prediction model to determine the predicted fault type of the server. The real-time data collected by the server BMC (Baseboard Management Controller) is connected to the fault prediction model, and the collected real-time operation data is parsed and formatted, and standardized into the input format required by the fault prediction model. At the same time, the real-time operation data is preprocessed, including missing value filling, outlier processing, and feature standardization, to ensure data quality and consistency.

[0076] The deployed fault model is linked to the data interface to continuously monitor the operating status of the server memory and hard disk. The newly collected real-time operating data is dynamically input into the fault prediction model for analysis and outputs the prediction results.

[0077] For example, when the failure probability of a server predicted by the fault prediction model exceeds a failure threshold, a fault warning is triggered, and the failure threshold may be 0.8. For particularly critical servers or devices, a lower threshold may be set to increase sensitivity.

[0078] Update the server's operating health status in real time to form a dynamic status monitoring chart for administrators to view at any time. Adjust the model output according to the latest data to improve the timeliness and reliability of the prediction results. At the same time, ensure the real-time nature of the system monitoring and prediction links, and ensure that fault signals can be captured and processed in the first time, thereby reducing system losses caused by delays.

[0079] S304: Generate fault diagnosis suggestions based on the predicted fault type.

[0080] For example, based on the predicted fault type, fault diagnosis suggestions are generated, such as checking specific components, replacing hardware, or optimizing load distribution, etc., which can help administrators take quick action. Traditionally, faults are often discovered after they have already occurred, which requires downtime for repair and may take a lot of time and cost. In this process, faults can be warned in advance, allowing maintenance personnel to take timely measures to reduce the time and cost required for maintenance, while avoiding business losses caused by long downtime.

[0081] Furthermore, the failure probability and failure level corresponding to the predicted failure type of the server can be determined based on the hard disk failure prediction characteristics and memory failure prediction characteristics. When the failure probability is greater than or equal to the probability threshold, an alarm signal with a failure level is generated. The alarm signal is used to instruct the user to allocate maintenance resources to the server according to the failure level.

[0082] For example, when the fault prediction model detects that the server memory or hard disk has a risk of failure, the early warning mechanism is activated, an alarm message can be generated in the BMC, and an alarm email push notification can be sent to the administrator's monitoring platform. At the same time, according to the fault risk level (low, medium, high), graded warnings are provided to help administrators reasonably allocate maintenance resources. The details of each warning (time, device, predicted value, risk level) are recorded in the system log for subsequent analysis and auditing.

[0083] The fault prediction method provided in the embodiment of the present application can timely detect potential memory and hard disk faults, thereby avoiding server system downtime or service interruption caused by unpredicted faults. This helps to reduce direct economic losses and reputation damage caused by production interruption, data loss and service interruption due to faults.

[0084] Furthermore, in the case of multiple servers in a production environment, by real-time monitoring of the status of each server and predicting possible failures, system administrators can make targeted resource allocation and optimization. For example, servers that may fail can be removed from the load balancing pool in a timely manner to prevent them from affecting the performance of other servers, thereby improving the operating efficiency and stability of the entire server cluster.

[0085] The above mainly introduces the solution provided by the embodiment of the present application from the perspective of the working principle of the baseboard management controller. It is understandable that in order to realize the above functions, the baseboard management controller includes hardware structures and / or software modules corresponding to the execution of each function. Those skilled in the art should easily realize that, in combination with the algorithm steps of each example described in the embodiments disclosed herein, the present application can be implemented in the form of hardware or a combination of hardware and computer software. Whether a function is executed in the form of hardware or computer software driving hardware depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0086] The embodiment of the present application can divide the functional modules of the baseboard management controller according to the above method example. For example, each functional module can be divided according to each function, or two or more functions can be integrated into one processor. The above integrated module can be implemented in the form of hardware or software functional module.

[0087] It should be noted that the division of modules in the embodiments of the present application is schematic and is only a logical function division. There may be other division methods in actual implementation. Figure 4 FIG. 2 shows a possible schematic diagram of the fault prediction device involved in the above-mentioned embodiments. Figure 4 As shown, the fault prediction device 400 may include: an acquisition module 401 , a determination module 402 and a generation module 403 .

[0088] The acquisition module 401 is used to support the fault prediction device 400 provided in the embodiment of the present application to perform the following steps: Figure 3 S301 in the fault prediction method shown.

[0089] The determination module 402 is used to support the fault prediction device 400 provided in the embodiment of the present application to perform the following Figure 3 S302 and S303 in the fault prediction method shown.

[0090] The generation module 403 is used to support the fault prediction device 400 provided in the embodiment of the present application to perform the following steps: Figure 3 S304 in the fault prediction method shown.

[0091] In a possible implementation, the device is specifically used to obtain historical operation data of a server. The historical operation data carries a fault type label. The historical operation data includes historical operation parameters of hardware devices, historical log data of hardware devices, historical hardware data of hardware devices, and historical software data of software systems. According to the historical operation parameters, historical log data, historical hardware data, and historical software data, the fault characteristics of the server are determined. The fault characteristics include hard disk fault characteristics and memory fault characteristics. According to the hard disk fault characteristics and memory fault characteristics, a fault prediction model is trained.

[0092] In a possible implementation, hard disk failure characteristics include SMART data characteristics, read / write speed characteristics, and run time characteristics. Memory failure characteristics include memory statistical characteristics, time series characteristics, and frequency domain characteristics. The device is specifically used to filter out SMART parameters, read / write speed parameters, run time parameters, memory statistical parameters, time series parameters, and frequency domain parameters from historical run parameters, historical log data, historical hardware data, and historical software data. According to SMART parameters and SMART change trends of SMART parameters in a preset time period, SMART data characteristics are determined. According to read / write speed parameters and read / write speed change trends of read / write speed parameters in a preset time period, read / write speed characteristics are determined. According to run time parameters, run time characteristics are determined. Statistical processing is performed on memory statistical parameters within a preset time period to obtain memory statistical characteristics. Feature extraction is performed on time series parameters based on a sliding window to obtain time series characteristics. Fourier transform is performed on frequency domain parameters to obtain frequency domain characteristics.

[0093] A possible implementation method, the device is specifically used to construct a sample set using the historical operation data of the server. Configure the number of decision trees of the fault prediction model, the depth of each decision tree, the number of features used by the nodes on each decision tree, the iteration termination condition of the fault prediction model, the number of samples of each node and the information gain of each node. Extract training samples from the sample set with replacement to form a training set. Use the training set to start training from the root node of each decision tree until all nodes of the decision tree are trained to obtain a fault prediction model. The root node is any node on the decision tree.

[0094] In a possible implementation, the device is specifically used to extract test samples from a sample set to form a test set. The test set is input into a trained fault prediction model to obtain a fault test result. The fault test result is evaluated using a Gini parameter to obtain a Gini index. Based on the Gini index, the parameters of the fault prediction model are adjusted.

[0095] In a possible implementation, the device is specifically used to fill missing values ​​in the real-time operation data using an interpolation method, and / or to remove or correct abnormal values ​​in the real-time operation data, and / or to remove duplicate values ​​in the real-time operation data.

[0096] In a possible implementation, the device is specifically used to determine the failure probability and failure level corresponding to the predicted failure type of the server based on the hard disk failure prediction characteristics and the memory failure prediction characteristics. When the failure probability is greater than or equal to the probability threshold, an alarm signal with the failure level is generated. The alarm signal is used to instruct the user to allocate maintenance resources to the server according to the failure level.

[0097] It should be noted that all relevant contents of each step involved in the above method embodiment can be referred to the functional description of the corresponding functional module and will not be repeated here.

[0098] The fault prediction device 400 provided in the embodiment of the present application is used to perform the above Figure 3 The fault prediction method shown can therefore achieve the same effect as the above-mentioned fault prediction method.

[0099] The present application also provides a baseboard management controller, which can be the baseboard management controller described in the above embodiments. Figure 1 The baseboard management controller 100 shown can execute the fault prediction method and related steps in the above method embodiment.

[0100] An embodiment of the present application also provides a computer-readable storage medium having instructions stored thereon, which, when executed, execute the fault prediction method and related steps in the above method embodiment.

[0101] The embodiment of the present application also provides a computer program product. When the computer program product is run on a computer, the computer executes the fault prediction method and related steps in the above method embodiment.

[0102] In some embodiments, the methods described herein may be implemented as computer program instructions encoded in a machine-readable format on a computer-readable storage medium or encoded on other non-transitory media or articles of manufacture.

[0103] The present application also provides a fault prediction system 200, such as Figure 5 As shown, the fault prediction system 200 includes at least one processor 501 and at least one interface circuit 502 .

[0104] As an example, when the fault prediction system 200 includes a processor and an interface circuit, the processor may be Figure 5The processor 501 shown in the solid line frame (or the processor 501 shown in the dotted line frame) may be Figure 5 The interface circuit 502 shown in the solid line frame (or the interface circuit 502 shown in the dotted line frame) is shown in the solid line frame. When the fault prediction system 200 includes two processors and two interface circuits, the two processors include Figure 5 The processor 501 shown in the solid line frame and the processor 501 shown in the dotted line frame, the two interface circuits include Figure 5 The interface circuit 502 shown in the solid line frame and the interface circuit 502 shown in the dotted line frame are not limited to this.

[0105] The processor 501 and the interface circuit 502 may be interconnected via a line. For example, the interface circuit 502 may be used to receive a signal. In another example, the interface circuit 502 may be used to send a signal to another device (e.g., the processor 501). For example, the interface circuit 502 may read a computer instruction stored in a memory and send the computer instruction to the processor 501. The processor 501 executes the instruction and implements the various steps in the above embodiments in combination with the input and output devices, such as implementing Figure 3 The various steps performed in the method embodiment shown. Of course, the fault prediction system 200 may also include other discrete components, which is not specifically limited in the embodiment of the present application.

[0106] Through the description of the above implementation methods, technical personnel in the relevant field can clearly understand that for the convenience and simplicity of description, only the division of the above-mentioned functional modules is used as an example. In actual applications, the above-mentioned functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above.

[0107] In the several embodiments provided in the present application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the modules or units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another device, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0108] The units described as separate components may or may not be physically separated, and the components shown as units may be one physical unit or multiple physical units, that is, they may be located in one place or distributed in multiple different places. Some or all of the units may be selected according to actual needs to achieve the purpose of the present embodiment.

[0109] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0110] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a readable storage medium. Based on this understanding, the technical solution of the embodiment of the present application is essentially or the part that contributes or all or part of the technical solution can be embodied in the form of a software product, which is stored in a storage medium, including several instructions to enable a baseboard management controller (which can be a single-chip microcomputer, chip, etc.) or a processor (processor) to perform all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read only memory (ROM), random access memory (RAM), disk or optical disk and other media that can store program code.

[0111] The above contents are only specific implementation methods of the present application, but the protection scope of the present application is not limited thereto. Any changes or substitutions within the technical scope disclosed in the present application shall be included in the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the protection scope of the claims.

Claims

1. A fault prediction method, characterized in that: Applied to a baseboard management controller, the baseboard management controller is used to monitor and manage a server, the server includes a hardware device, and a software system runs on the server, the method includes: Acquire real-time operation data of the server; the real-time operation data includes operation parameters of the hardware device, log data generated by the hardware device during operation, hardware data of the hardware device and software data of the software system; Determining the fault prediction characteristics of the server according to the operating parameters, the log data, the hardware data and the software data; the fault prediction characteristics include hard disk fault prediction characteristics and memory fault prediction characteristics; Inputting the hard disk fault prediction feature and the memory fault prediction feature into the trained fault prediction model to determine the predicted fault type of the server; Based on the predicted fault type, a fault diagnosis suggestion is generated.

2. The method according to claim 1, characterized in that The training process of the fault prediction model includes: Acquire historical operation data of the server; the historical operation data carries a fault type label; the historical operation data includes historical operation parameters of the hardware device, historical log data of the hardware device, historical hardware data of the hardware device and historical software data of the software system; Determining the fault characteristics of the server according to the historical operating parameters, the historical log data, the historical hardware data and the historical software data; the fault characteristics include hard disk fault characteristics and memory fault characteristics; The fault prediction model is trained according to the hard disk fault characteristics and the memory fault characteristics.

3. The method according to claim 2, characterized in that The hard disk fault characteristics include SMART data characteristics, read and write speed characteristics and running time characteristics; the memory fault characteristics include memory statistical characteristics, time series characteristics and frequency domain characteristics; Determining the fault characteristics of the server according to the historical operating parameters, the historical log data, the historical hardware data, and the historical software data includes: Filter out SMART parameters, read / write speed parameters, run time parameters, memory statistics parameters, time series parameters and frequency domain parameters from the historical operating parameters, the historical log data, the historical hardware data and the historical software data; Determining the SMART data feature according to the SMART parameter and the SMART change trend of the SMART parameter in a preset time period; Determining the reading and writing speed characteristics according to the reading and writing speed parameters and the reading and writing speed change trends of the reading and writing speed parameters in a preset time period; determining the runtime characteristic according to the runtime parameter; Performing statistical processing on the memory statistical parameters within a preset time period to obtain the memory statistical features; Extracting features of the time series parameters based on a sliding window to obtain the time series features; Performing Fourier transform on the frequency domain parameters to obtain the frequency domain features.

4. The method according to claim 2, characterized in that: The training of the fault prediction model according to the hard disk fault characteristics and the memory fault characteristics includes: Building a sample set using historical operation data of the server; Configure the number of decision trees of the fault prediction model, the depth of each decision tree, the number of features used by each node on the decision tree, the iteration termination condition of the fault prediction model, the number of samples of each node and the information gain of each node; Extracting training samples from the sample set with replacement to form a training set; The training set is used to start training from the root node of each decision tree until all nodes of the decision tree are trained to obtain the fault prediction model; the root node is any node on the decision tree.

5. The method according to claim 4, characterized in that The method further comprises: Extracting test samples from the sample set to form a test set; Inputting the test set into the trained fault prediction model to obtain a fault test result; The fault test result is evaluated using the Gini parameter to obtain the Gini index; Based on the Gini index, the parameters of the fault prediction model are adjusted.

6. The method according to claim 1, characterized in that The method further comprises: Filling missing values ​​in the real-time operation data using interpolation; and / or, Eliminate or correct abnormal values ​​in the real-time operation data; and / or, The duplicate values ​​in the real-time operation data are removed.

7. The method according to claim 1, characterized in that The method further comprises: Determining a failure probability and a failure level corresponding to a predicted failure type of the server according to the hard disk failure prediction feature and the memory failure prediction feature; When the failure probability is greater than or equal to a probability threshold, an alarm signal with the failure level is generated; the alarm signal is used to instruct a user to allocate maintenance resources to the server according to the failure level.

8. A fault prediction device, characterized in that: Applied to a baseboard management controller, the baseboard management controller is used to monitor and manage a server, the server includes a hardware device, a software system runs on the server, and the device includes: An acquisition module, used to acquire real-time operation data of the server; the real-time operation data includes operation parameters of the hardware device, log data generated by the hardware device during operation, hardware data of the hardware device and software data of the software system; A determination module, configured to determine the fault prediction characteristics of the server according to the operating parameters, the log data, the hardware data and the software data; the fault prediction characteristics include hard disk fault prediction characteristics and memory fault prediction characteristics; input the hard disk fault prediction characteristics and the memory fault prediction characteristics into the trained fault prediction model to determine the predicted fault type of the server; A generating module is used to generate a fault diagnosis suggestion based on the predicted fault type.

9. A baseboard management controller, characterized in that: The baseboard management controller includes a processor and a memory, wherein the memory stores computer executable instructions that can be executed by the processor, and the processor executes the computer executable instructions to implement the fault prediction method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the fault prediction method according to any one of claims 1 to 7 is implemented.

Images